4 ms·

Out of interest, when the chemically accurate simulations are made are there any surprises? Or do the estimations that are normally used good enough for most pu

by papaf 14y ago

Out of interest, when the chemically accurate simulations are made are there any surprises? Or do the estimations that are normally used good enough for most purposes?

escherba 14y ago
I'm not an expert on Comp Chem, but you might want to rephrase your question. If the simulation is "chemically accurate", of course it will match the real-world, by definition of "accurate"...
- papaf 14y ago
  My understanding of these things is that the dynamics of the real world are not easily measurable at this scale. An alternative is to simulate using using packages such as Gromacs: http://www.gromacs.org/About_Gromacs These packages are truly incredible but use relatively crude approximations and are in wide use. It is possible to get a decent paper out that uses a simulation as evidence to support an idea. The aim of my question was to see if chemically accurate simulations come up with significantly different answers or to see if the current way of doing things is a good enough approximation.
  - hagy 14y ago
    You bring up an excellent point, papaf. The accuracy of the empirical force fields used in molecular dynamics simulation engines (such as Gromacs) is a hotly debated topic. Even without quantum computing the issue can be addressed with conventional computers that perform quantum mechanical calculations on interacting molecular fragments. The forces calculated from quantum mechanics can then be compared with those calculated from molecular dynamics force fields (as an example see Sherrill et al. http://onlinelibrary.wiley.com/doi/10.1002/jcc.21226/full http://onlinelibrary.wiley.com/doi/10.1002/jcc.21226/full ). Such studied have shown that current force fields fail to model certain chemical interactions and need improved. Specifically, the underlying functional forms used to model molecular forces need revised. Currently such investigations are limited in scope by the large computational resources required to perform a single quantum mechanical calculation on a molecular fragment. With quantum computers, tens of thousands of such calculations could be performed and the results could be used to optimize new molecular force fields through multivariate regression.
    - Xcelerate 14y ago
      Nice to see another GT person on HN! (Did my undergrad there.) Have you worked with Dr. Sherrill? It's funny you mention him; I was actually reading one of his presentations on electron-electron correlation last night.
      - hagy 14y ago
        Cool. I'm a grad student in another chemistry theory group. Dr. Sherrill definitely has the best notes on electronic structure calculations.
adrianhoward 14y ago
Out of interest, when the chemically accurate simulations are made are there any surprises? Or do the estimations that are normally used good enough for most purposes? Well - it's more that there are entire classes of things that we can't simulate in any reasonable amount of time that (in theory) QCs should be very good at. Protein folding for example. Finding the lowest energy state that protein's fold into is really, really hard and slow. QCs can theoretically do it very, very quickly. This opens up whole areas of experimentation and validation that are closed to us at the moment because the feedback cycle on solutions is so darn slow and/or inaccurate. Protein folding errors are at the heart of diseases like Alzheimer’s, Huntington’s and Parkinsons. QCs capable of simulating the chemistry involved would be a huge help in attacking those problems.
asafira 14y ago
D.E. Shaw current is one of the leaders (if not the leader) in molecular dynamics (i.e., studying the motion of the specific atoms in a molecule, or watching the dynamics of a chemical reaction), and it's a really hard problem. Roughly speaking, you have to calculate a constant (but also fairly large) number of operations to calculate the configuration of the molecule(s) for every _femtosecond_ of time. It would be great to see things happen at a tens of microseconds, but that is a LOT of computation. This is useful becaue it would mean that we could see exactly how reactions take place and perhaps even engineer interesting chemical/biological phenomena. Certain processes in your body depend on proteins moving certain substances or reacting in certain ways, and if we can simulate all that with a computer, we can start to build chemical/biological tools for, for example, fighting certain diseases

by papaf 14y ago

Out of interest, when the chemically accurate simulations are made are there any surprises? Or do the estimations that are normally used good enough for most purposes?

escherba 14y ago

I'm not an expert on Comp Chem, but you might want to rephrase your question. If the simulation is "chemically accurate", of course it will match the real-world, by definition of "accurate"...

papaf 14y ago
My understanding of these things is that the dynamics of the real world are not easily measurable at this scale. An alternative is to simulate using using packages such as Gromacs: http://www.gromacs.org/About_Gromacs These packages are truly incredible but use relatively crude approximations and are in wide use. It is possible to get a decent paper out that uses a simulation as evidence to support an idea. The aim of my question was to see if chemically accurate simulations come up with significantly different answers or to see if the current way of doing things is a good enough approximation.
- hagy 14y ago
  You bring up an excellent point, papaf. The accuracy of the empirical force fields used in molecular dynamics simulation engines (such as Gromacs) is a hotly debated topic. Even without quantum computing the issue can be addressed with conventional computers that perform quantum mechanical calculations on interacting molecular fragments. The forces calculated from quantum mechanics can then be compared with those calculated from molecular dynamics force fields (as an example see Sherrill et al. http://onlinelibrary.wiley.com/doi/10.1002/jcc.21226/full http://onlinelibrary.wiley.com/doi/10.1002/jcc.21226/full ). Such studied have shown that current force fields fail to model certain chemical interactions and need improved. Specifically, the underlying functional forms used to model molecular forces need revised. Currently such investigations are limited in scope by the large computational resources required to perform a single quantum mechanical calculation on a molecular fragment. With quantum computers, tens of thousands of such calculations could be performed and the results could be used to optimize new molecular force fields through multivariate regression.
  - Xcelerate 14y ago
    Nice to see another GT person on HN! (Did my undergrad there.) Have you worked with Dr. Sherrill? It's funny you mention him; I was actually reading one of his presentations on electron-electron correlation last night.
    - hagy 14y ago
      Cool. I'm a grad student in another chemistry theory group. Dr. Sherrill definitely has the best notes on electronic structure calculations.

adrianhoward 14y ago

Out of interest, when the chemically accurate simulations are made are there any surprises? Or do the estimations that are normally used good enough for most purposes? Well - it's more that there are entire classes of things that we can't simulate in any reasonable amount of time that (in theory) QCs should be very good at. Protein folding for example. Finding the lowest energy state that protein's fold into is really, really hard and slow. QCs can theoretically do it very, very quickly. This opens up whole areas of experimentation and validation that are closed to us at the moment because the feedback cycle on solutions is so darn slow and/or inaccurate. Protein folding errors are at the heart of diseases like Alzheimer’s, Huntington’s and Parkinsons. QCs capable of simulating the chemistry involved would be a huge help in attacking those problems.

asafira 14y ago

D.E. Shaw current is one of the leaders (if not the leader) in molecular dynamics (i.e., studying the motion of the specific atoms in a molecule, or watching the dynamics of a chemical reaction), and it's a really hard problem. Roughly speaking, you have to calculate a constant (but also fairly large) number of operations to calculate the configuration of the molecule(s) for every _femtosecond_ of time. It would be great to see things happen at a tens of microseconds, but that is a LOT of computation. This is useful becaue it would mean that we could see exactly how reactions take place and perhaps even engineer interesting chemical/biological phenomena. Certain processes in your body depend on proteins moving certain substances or reacting in certain ways, and if we can simulate all that with a computer, we can start to build chemical/biological tools for, for example, fighting certain diseases